Identifying Children Metacognitive Monitoring Performance Through Facial Expressions
summary
Built a labeled video dataset called Affect2Metacognition from recordings of children using the Meta-Brainhood app with six puzzle types and post-round confidence judgments, each clip paired with a metacognitive monitoring score. Neural networks trained on the dataset classify overconfident versus underconfident responses at about 60–70% accuracy, with model performance falling when subject diversity is low. Proposes real-time scaffolding based on detected MMPs and plans to test impact on learning outcomes.
AI-written summary · about
related talks
20mEmpowering Education and Therapy with Data Science
10mMIND MELD: Personalized Meta-Learning for Robot-Centric Imitation Learning
15mRecognizing Emotional Expression in Game Streams
49mAffective Computing – Bringing Humans and Machines Closer Through Emotions
07mAssessing Cognitive Performance Using Physiological and Facial Features: Generalizing Across ...
34mEnhancing Customer Experiences with Emotional AI,
11mAmbient Intelligence and Emotionally Intelligent Devices
23mPrediction of music pairwise preferences from facial expressions